OpenAI engineering lead Thibault Sottiaux disclosed on May 24 that a recently shipped optimization had reduced cache hit rates during context compaction in long-running Codex sessions, causing usage limits to deplete faster than intended. The team identified and rolled back the change, then reset usage limits across all accounts. The post drew over 648,000 views, making it the most-viewed AI story on the day.
The incident is at least the ninth time Sottiaux has announced a Codex usage limit reset since December 2025, according to community tracking — a pattern that reflects both the pace of iteration on Codex and the sensitivity of its user base to quota changes. Reaction to the fix was broadly positive, with over 80% of sentiment-tracked replies expressing relief or excitement at restored access heading into the weekend.